DiPhyx Advances Molecular Dynamics with Enhanced GROMACS GPU Scalability


DiPhyx, utilizing its cutting-edge infrastructure on GCP and self-hosted servers (managed by dxflow), proudly integrates the latest enhancements in GROMACS, the foremost software for molecular dynamics simulations. With the recent update, GROMACS now supports multi-node NVIDIA GPU scalability, notably enhancing the software's ability to analyze molecular behaviors critical for drug discovery and molecular research.

Enhanced GPU Particle Mesh Ewald (PME) Decomposition

The highlight of the GROMACS 2023 release is the enablement of GPU Particle Mesh Ewald decomposition, which significantly improves the software’s performance on NVIDIA GPUs available through GCP and DiPhyx’s hybrid environments. This update facilitates up to a 21x increase in performance by allowing GPU direct communication, revolutionizing how molecular dynamics simulations are processed.

Improved Multi-Node Performance

Traditionally, GROMACS faced scalability challenges due to its single-GPU dependency for PME long-range force calculations. The latest enhancements address these limitations by decomposing PME calculations across multiple GPUs. This decomposition, enabled by NVIDIA's cuFTTMp library and NVSHMEM for fast inter-GPU communication, allows DiPhyx users to effectively scale their simulations across more nodes without encountering the typical bottlenecks.

Real-World Performance Results

In practical terms, the improvements are substantial:

  • STMV System Simulation: Initial tests showed a 2x increase in single-node performance. With PME decomposition, performance scales efficiently, showing more than double the performance at 4 nodes, and tripling at 8 nodes (32 GPUs).
  • BenchPEP System Simulation: For larger molecular dynamics simulations, the need to distribute PME across multiple GPUs is crucial. This setup shows exceptional scalability, peaking at 21x faster than legacy setups at 64 nodes.

Setting Up GROMACS on DiPhyx Platforms

To leverage these advancements in GROMACS within DiPhyx's infrastructure:

  1. Create a GPU enabled Compute-Unit, currently avaiable through
    1. GCP
    2. "Self": which stands for self-hosted server. See the instruciton at dxflow.ai
  2. Select the GPU enabled GROMACS through our software page

Two step, couple of click is all you need to do!


The enhancements in GROMACS 2023 dramatically extend the capabilities of executing multi-node GPU clusters, facilitating more complex molecular dynamics problems to be solved swiftly. This leads to faster time-to-market solutions and a deeper, more comprehensive analysis of molecular interactions.

Leverage the power of DiPhyx's advanced computational resources and the new GROMACS features to accelerate your research and drive discovery to unprecedented levels. For further details on utilizing GROMACS within DiPhyx. Visit our website or contact us directly at info@diPhyx.com.

Contact us
+1 (619) 693-6161
Follow us on
@2023-2024 DiPhyx, Inc.